In only a handful of years, synthetic intelligence (AI) has reshaped how we expect and have interaction with the world. Generative AI has already reworked our digital habits, changing into for a lot of a major, one-stop supply of knowledge. It’s remapping the office, the classroom, the halls of politics and even the material of tradition itself.
“AI didn’t simply change know-how; it modified human habits. And as soon as human habits adjustments, society strikes at a wholly completely different velocity,” mentioned Simon Lee, CEO of Flitto, an AI-powered translation and dwell interpretation supplier. His remarks got here Tuesday night on the Qatari ambassador’s residence in Seoul, in the course of the Corea Picture Communication Establishment’s (CICI) Korea CQ Discussion board.
In the present day, the world has moved past fascination with the know-how’s promise. Leaders now need their corporations to stake early claims within the international AI race, with the hope of harnessing its energy. The competition to develop and command AI is not looming, however is absolutely underway.
This race spans each layer of know-how, Lee defined: {hardware}, algorithms and knowledge.
On the floor stage is {hardware}, such because the chips that make complicated AI fashions potential. Graphics processing items, or GPUs, allow AI to carry out large calculations concurrently. They’re what drove Nvidia to change into a chip making big valued at $5 trillion right now.
“International locations noticed Nvidia’s success, and they’re now attempting to develop their very own variations of this know-how — not full GPUs, however smaller, extra environment friendly chips known as NPUs,” he mentioned. These NPUs can run smaller AI fashions on compact units like telephones with larger velocity and a fraction of power consumption, whereas GPUs stay important for powering big fashions.
In Korea, two startups — Rebellions and Furiosa AI — have emerged as robust contenders on this race.
As massive AI fashions have swept throughout the globe, a brand new concern has surfaced: many nations now rely upon large techniques constructed by two superpowers, the USA and China.
“There may be all the time the worry that consumer knowledge may stream again to these nations,” the CEO mentioned. “And since these fashions are skilled inside a selected tradition and worth system, their solutions naturally include cultural or political bias.”
That’s the reason many governments try to construct their very own nationwide AI fashions that replicate their languages, cultures, social norms and geopolitical realities.
“In brief, nations don’t simply wish to use AI anymore. They wish to form it, practice it on their very own knowledge and ensure it displays their very own values and nationwide priorities,” he added.
However the actuality is that not each nation can develop its personal basis mannequin, which calls for staggering computing energy, billions of information factors and immense monetary sources.
For that purpose, some nations, together with Korea, are shifting their focus. Somewhat than pouring all sources into big basis fashions, they’re investing in vertical AI, or techniques tailor-made to particular industries and use circumstances.
“We could not invent each foundational know-how, however we are able to construct specialised purposes and companies. And that is precisely why vertical AI is changing into so necessary,” Lee mentioned, noting that Flitto’s personal real-time multilingual AI interpretation service, designed particularly for worldwide conferences, is one such instance.
Lastly, there’s uncooked knowledge, which Lee calls “the one component that really determines whether or not an AI succeeds or fails, but unusually, the one factor folks speak concerning the least.”
The availability of high-quality, human-generated knowledge is shrinking quick. Some consultants estimate that by 2027, we may have exhausted almost all digital knowledge out there for AI coaching. What’s going to we feed AI within the years to come back, if it has already consumed the majority of humanity’s information? What stays earlier than it begins cannibalizing its personal outputs and collapsing beneath the burden of recycled knowledge?
It’s an ongoing drawback with no clear resolution, however the CEO argued that this knowledge scarcity ought to reframe how we take into consideration AI improvement. “The aim of constructing AI is to generate new knowledge from human beings in order that we are able to practice that knowledge into our options,” he famous. “Flitto offers real-time interpretation for folks. Certain, it helps break down language limitations, however our most important aim is to gather large quantities of human pronunciation, accents [and speech patterns] coming from every kind of individuals.”
That rising reservoir of information can then feed different AI-powered companies, from voice recognition to machine translation, that folks work together with of their day by day lives.
